udf in parallel processing
i have built an udf for serial processing and it works, now i would like to make run this function also in the parallel processing but i know that there's something to do, some steps to follow to do this.
Can someone tell me what is the procedure?
Thanks a lot
You need to first read about "Parallelizing your UDF". Try Google first.
To sum up a little about what is going on, there are two terms you need to be aware of: Host and Node.
The Host is basically where the cortex (or Fluent GUI) is running and the Nodes are the processors where the calculations are occurring. There are no calculations happening on the Host. You rewrite your UDF to recognize these two new concepts and Fluent will compile two sets of your UDF (Host and Node). If you program your UDF correctly, you should be able to use the parallel UDF in serial as the compiler directives will be ignored.
You will need to keep in the back of your mind the answers to the following questions:
What part of my serial UDF do calculations? (This will be for your nodes)
Where am I defining my variables? (If defined in host, it will not be in nodes)
Do I need to pass any data in between nodes? (You will need to use macros like node_to_host)
Obviously this is just a real short description on parallelizing your UDF, but it at least introduces you to some of the vocabulary to search for.
|All times are GMT -4. The time now is 07:23.|